The invention relates to a method for producing a preferably asymmetrical lens element from a tempered blank; to a lens blank for microlithography, preferably with cylindrical geometry; as well as to lens elements and to a projection lens with such a lens element.
Lens elements made of fused silica are, for example, used in projection exposure apparatuses for microlithography. In such apparatuses, radiation generated by a usually pulsed laser at an operating wavelength of e.g. 248 nm (KrF laser) or 193 nm (ArF laser) is imaged to a sharply delimited very homogeneously illuminated image field by means of an illumination system, in which image field a mask is arranged. A pattern that is provided on the mask is imaged, by means of a downstream projection lens, at a reduced scale on a semiconductor wafer that comprises a light-sensitive layer.
For wavelengths of 250 nm and below, which wavelengths are used in such systems, birefringence of the fused silica material plays an important role. The term “birefringence” refers to the splitting, which occurs in optically anisotropic materials, of the incident radiation into two partial beams that are polarized perpendicularly in relation to each other and in relation to the direction of propagation (ordinary and extraordinary beams) of different propagation speeds. The axis with the higher propagation speed is also referred to as the “fast axis”.
As a result of the different propagation speeds, after passing through the optical material the two partial rays undergo a phase shift, which in an imaging optical arrangement can have a negative effect on its ability to provide a true image, i.e. on the imaging contrast. Therefore, optical components used in lithography optics applications should have as little birefringence as possible. Furthermore, in so-called polarization-preserving lithography systems a polarization state, once set in the illumination system, should be preserved as well as possible right to the wafer, i.e. the projection lens should maintain polarization to the greatest possible extent, which is made considerably more difficult by birefringence.
The blanks made of synthetic fused silica that are used in the manufacture of lenses, which blanks usually consist of cylindrical discs, are manufactured by way of flame hydrolysis (soot process) or in the direct process (direct vitrification) at high temperatures. In order to prevent birefringence, which can occur as a result of mechanical strain during fast cooling of the blanks, the blanks are subjected to tempering treatment, i.e. for an extended period of time (e.g. 50 h) they are held at high temperatures (usually in excess of 1800° C.) before they are slowly cooled to room temperature.
DE 10 2004 009 577 A1 describes a method for manufacturing optical components, in which method a first tempering treatment at high temperatures is followed by a second tempering treatment at low temperatures, e.g. between 350° C. and 800° C. By means of the second tempering treatment the essentially tangential alignment of the fast axis of the birefringence on the longitudinal axis of the cylinder should be able to be transformed into an essentially radial alignment. Furthermore, the blanks manufactured in this way are said to be more resistant to decompacting (rarefaction).
In order to determine the strain birefringence (SBR) of the lenses cut from the blanks before they are installed in an optical system, and if necessary to be able to initiate measures to compensate for it, the strain birefringence of each blank is measured after tempering, namely along the longitudinal axis of the cylindrical blank-disc (z-direction), which longitudinal axis essentially corresponds to the direction of passage of the light. In this method, a value of the strain birefringence, which value has been integrated over the z-axis and has been averaged, is determined. For the purpose of measuring, devices are used that produce a strain birefringence at 633 nm (He—Ne laser) and that can scan the blank automatically in the x- and y-directions. In most cases involving fused silica, an essentially rotationally-symmetrical distribution of the SBR in the x-y-plane is detected in this process, wherein its absolute value increases squarely with the distance from the center point (corresponding to the longitudinal axis of the cylinder). In this arrangement the orientation of the fast axis of the SBR normally is predominantly tangential or radial. The specified SBR at 633 nm in the case of averaging across the circumference of the optically clear diameter is typically in an interval of between 0.2 nm/cm and 1 nm/cm, most often at 0.5 nm/cm.
When using the lenses, formed from the blanks, in optical systems of the applicant, an SBR has been observed that differs from the SBR measured on the blanks by means of the above methods, even after the contributions that material processing, material refining and mounting techniques make to SBR have been taken into account. During investigations as to the reasons for this material-related SBR, the inventor found that in particular highly curved lenses, already after cutting the lens-form into the blank and after polishing, but before mounting and coating, have an SBR that is higher than the SBR predicted on the basis of initial measurement of the blank by weighting with the local lens density. Furthermore, it has been observed that a tangential distribution of the fast axis in the blank can transform into a radial distribution in the lens (or vice-versa).
Such a difference between the SBR of the cut lens and the SBR measured on the blank has been observed in particular in the case of asymmetrical lenses. In an asymmetrical lens the radii of curvature of the two optically effective surfaces differ as far as their absolute values and/or their signs are concerned. In aspherical lenses where if applicable no radii of curvature have been defined, the term “asymmetrical lens” refers to a lens in which no plane can be determined, in relation to which the lens has a mirror symmetry. The difference described above is particularly pronounced in the case of highly curved lenses, i.e. in lenses in which the two radii of curvature differ significantly from each other, e.g. in an extreme case in plano-convex lenses, but also in meniscus lenses that essentially have the same large radii of curvature with opposite signs.
In order to understand the origin of the effects described above, the density distribution of the blank has to be examined in more detail, which density distribution results during the tempering process. As stated above, during tempering, the blank is heated to a maximum temperature of up to 1800° C. (glass temperature), is held for several hours to days at this temperature, and is then slowly cooled at a defined rate. Generally speaking, the slower the rate of cooling, the higher the resulting density. Furthermore, there is a temperature range of 1000° C. to 1500° C. in which there is an anomaly in the dependence of the density on the rate of cooling; i.e. in this temperature range the density increases as the cooling rate increases. By controlling the cooling rate it is possible to have an influence as to which process dominates. Apart from this, the OH content and thus the coefficient of thermal expansion (CTE) of the blank can also have a radial dependence, and can thus also lead to a rotationally symmetrical density distribution.
In the context of the density distribution arising during the tempering process it is essential that normally the slower the rate of cooling, the higher the resulting density. Since cooling takes place over the surfaces of the cylindrical blank, volume elements that are close to the edge cool faster than do volume elements that are near the centre, and therefore have a different, usually lower, density.
FIG. 5a shows a lateral view (zx-section of an xyz-coordinate system) of a tempered blank 1, during whose tempering the cooling-rate anomaly did not dominate, and which blank 1 showed adequately homogeneous OH distribution. The near-center volume elements of said blank 1 therefore have a higher density than its edge regions. The regions 2a to 2d, which are shown in a dotted line in FIG. 5, show regions of identical density. They are nested in the manner of onion skins and in the center form spheroids (regions 2a, 2b), while towards the edge they extend to the corners (regions 2c, 2d), i.e. they tend to become cylindrical discs. Overall, the tempered blank 1 has a density distribution which extends rotation-symmetrically in relation to the z-axis as well as mirror-symmetrically in relation to a central plane (not shown) of the tempered blank 1, which central plane extends so as to be perpendicular in relation to the z-direction.
In the tempered blank 1 (tensile) strain 3a to 3d forms, which acts perpendicularly to the region 2a to 2d and whose amount and direction is shown by lines in FIG. 5a. The amount of the strain 3a to 3d, and thus the amount of the birefringence, increases in the tempered blank 1 from the inside to the outside.
In the case of SBR-measuring in z-direction in the manner described above, the strain components in the xy-plane are integrated along the z-direction over the thickness D of the tempered blank 1. Strain 3d that occurs parallel in relation to the z-axis is not detected, while strain 3a in circumferential direction is fully detected, which is consistent with the observed r2-distribution of the density amplitude. In relation to tension 3b, 3c with a 45° orientation, only the xy-component is detected, while the z-components are not detectable during standard measuring, and even during measuring with the blank tilted they are only detected conditionally, because the z-components of the strain 3b and 3c extending into the corners act in opposite directions, thus canceling each other out. If, as is the case in the state of the art, asymmetrical lens elements are cut from the tempered blank 1, as shown in FIG. 5b in relation to a meniscus lens element 4 and in FIG. 5c in relation to a plano-convex lens element 5, two effects occur, which are described in more detail below.
At first strain 3b occurs in the edge regions of the lens elements 4, 5, which strain extends so as to be essentially parallel in relation to the lens surfaces. Depending on the precise beam path, this strain can be almost perpendicular in relation to the direction of light and will thus result in high SBR observed. In contrast to this, in the middle, i.e. along the longitudinal axis of the tempered blank 17 the strain extends parallel in relation to the z-axis. As long as the beam path in this arrangement extends so as to be more or less parallel in relation to the z-axis, no SBR is observed. Only if the beam path extends obliquely through the middle of the lens, do SBR components occur; a situation which differs from that of e.g. a symmetrical biconvex lens.
Moreover, the volume of the tempered blank 1, from which the lens elements 4, 5 are formed, comprises a multitude of regions of equal density 2a to 2d. In the middle, i.e. along the longitudinal axis of the tempered blank 1, the density gradients extend parallel in relation to the lens surfaces, while at the edge they extend so as to be perpendicular to them. If the form of a body comprising internal mechanical strain is changed (in the present example from a cylindrical shape to a meniscus or piano-convex shape) then said body attempts to again assume a state of minimum energy. Said body will thus slightly deform in relation to the intended contour and in this process the strain will partly relax and partly shift. A body with high density gradients that are oriented differently, which body is formed by the lens elements 4, 5, thus behaves in an unfavorable manner in the sense that the shift in strain can be calculated in advance only with considerable difficulty and in that the precise effect of such shift in strain can only be clarified by means of experiments or by means of elaborate simulation.
If asymmetrical lens elements are cut from tempered blanks in a manner described in FIG. 5b and FIG. 5c, the unfavorable effects described above thus occur, namely on the one hand a high SBR observed in the edge regions of the lens elements, and on the other hand a shift in the strain after cutting the lens elements, which shift results in a shift of the SBR.